[SciPy-user] Trial run of object arrays returning the objects and *not* object array scalars

Fernando Perez Fernando.Perez at colorado.edu
Wed Jan 11 12:37:18 CST 2006

Travis Oliphant wrote:
> In SVN there is a version of numpy that does not return object array 
> scalars when using object arrays.  The machinery for object array 
> scalars is still there for now in case we decide that it is actually 
> useful (it's a simple change to go back to them).
> The down side is that object arrays now do not follow the rule that 
> items selected from them have the methods and attributes of arrays.  The 
> up-side is that now you don't have to use .item() to get to the actually 
> object you stored in the object array.
> Waiting to hear which side complains the loudest....

Thanks.  I do realize that this is a tricky balancing act between two opposing 
constraints, with good arguments to be made either way.  I just hope I don't 
find myself arguing for the opposite three months from now, because of having 
to special-case scalar access for 'O' arrays :)

Ah, I wish we could find some magical solution that would address all the 
issues in one clean shot.

This is certainly worth getting more feedback and testing from others; I am 
the first to admit I may only be seeing a small part of the issue.



More information about the SciPy-user mailing list